So basically you're using it as an AI mother in law, nice.

Of course, you're right. I didn't intend to criticise him.

I just think that it's not a good idea to transfer this megatrend to healthcare.

Yes, those systems can be highly accurate, but as we all have experienced, this is not a stable or consistent property. A very good model can give you an ingenious answer one minute and an utterly dumb one the next. My guess is that it's simply a consequence of the extremely high complexity of the 'plant' (natural language, language interfacing with the real world), leading to some highly non linear behaviour.

EDIT: I mean, those systems accumulated so much complexity around the attention based next token predictor.

Say i feed the largest LLM a book of an alien civilisation, that it definitely hasn't seen before. Then this tiny piece of text muds the vast ocean (latent space) of the model minimally. It will not be able to cite from that book reliably after that fine tuning. Especially for rare books, because them being rare implies, that there aren't 1000s of other books, that encode the same information.

It's general language modelling capabilities might get an iota better, of course. But for putting factual information into it, wouldn't RAG be a much more solid approach?

I had to lol at the doom scene.

Yes exactly. Enjoying the beauty of nature is (thankfully) completely decoupled from the job/money/society side of things.